STAR Development and Protocol Comparison Fisk, Liam M.; Le Compte, Aaron J.; Shaw, Geoffrey M. ...
IEEE transactions on bio-medical engineering/IEEE transactions on biomedical engineering,
12/2012, Letnik:
59, Številka:
12
Journal Article, Web Resource
Recenzirano
Odprti dostop
Accurate glycemic control (AGC) is difficult due to excessive hypoglycemia risk. Stochastic TARgeted (STAR) glycemic control forecasts changes in insulin sensitivity to calculate a range of glycemic ...outcomes for an insulin intervention, creating a risk framework to improve safety and performance. An improved, simplified STAR framework was developed to reduce light hypoglycemia and clinical effort, while improving nutrition rates and performance. Blood glucose (BG) levels are targeted to 80-145 mg/dL, using insulin and nutrition control for 1-3 h interventions. Insulin changes are limited to +3U/h and nutrition to ±30% of goal rate (minimum 30%). All targets and rate change limits are clinically specified and generalizable. Clinically validated virtual trials were run on using clinical data from 371 patients (39841 h) from the Specialized Relative Insulin and Nutrition Tables (SPRINT) cohort. Cohort and per-patient results are compared to clinical SPRINT data, and virtual trials of three published protocols. Performance was measured as time within glycemic bands, and safety by patients with severe (BG <; 40 mg/dL) and mild (%BG <; 72 mg/dL) hypoglycemia. Pilot trial results from the first ten patients (1486 h) are included to support the in-silico findings. In both virtual and clinical trials, mild hypoglycemia was below 2% versus 4% for SPRINT. Severe hypoglycemia was reduced from 14 (SPRINT) to 6 (STAR), and 0 in the pilot trial. AGC was tighter than both SPRINT clinical data and in-silico comparison protocols, with 91% BG within the specified target (80-145 mg/dL) in virtual trials and 89.4% in pilot trials. Clinical effort (measurements) was reduced from 16.2/day to 11.8/day (13.5/day in pilot trials). This STAR framework provides safe AGC with significant reductions in hypoglycemia and clinical effort due to stochastic forecasting of patient variation - a unique risk-based approach. Initial pilot trials validate the in-silico design methods and resulting protocol, all of which can be generalized to suit any given clinical environment.
•Stress, exercise and fatigue identified in type 1 diabetic glycaemic model data.•Identified parameter distribution adheres to 1/√n rule for increasing data.•Robust to long-term drift in metabolism, ...producing comparable results.•Successful identification of secondary effect parameters in sparse, noisy data.
Some individuals with type 1 diabetes mellitus find self-managed glycaemic control difficult due to the confounding influence of secondary effects. Stress and sleep deprivation temporarily lower insulin sensitivity (SI), often resulting in hyperglycaemia, while aerobic exercise depletes glucose, leading to hypoglycaemia if treatment is unchanged. This study tests the estimation of these factors and circadian rhythms of SI in noisy data. Sparse, irregular and noisy virtual blood glucose data, mimicking the glycaemic dynamics of an individual with type 1 diabetes, was created via adapted pharmacokinetic–pharmacodynamic models of glucose and insulin that included the impact of the secondary effects. A Gauss–Newton algorithm was used to recover the original model parameters for SI, stress, fatigue and exercise. During longer identification periods, compensation was made for drift in SI. Monte Carlo analyses were undertaken to validate the methods. The coefficient of variation (CV) in all parameters decreased as the data accumulated in proportion to the 1/n rule (R2 > 99.9%). Relatively small biases from the original parameter values occurred (<1%). Long term drift trends in SI were captured and did not obscure estimation of the secondary effects (biases < 1%, CV approximately equivalent to drift free outcomes). Adherence to the 1/n trend indicates a robust identification method and the ability of accumulating data to override the effect of measurement error. Compensation for SI drift allows viable observation of secondary effects and SI rhythms over longer time periods. Collectively, these outcomes indicate that quality results for identified parameters could be obtained during in vivo studies.
Clinical Validation of the Quick Dynamic Insulin Sensitivity Test Docherty, Paul D.; Berkeley, Juliet E.; Lotz, Thomas F. ...
IEEE transactions on bio-medical engineering/IEEE transactions on biomedical engineering,
05/2013, Letnik:
60, Številka:
5
Journal Article
Recenzirano
Odprti dostop
The quick dynamic insulin sensitivity test (DISTq) can yield an insulin sensitivity result immediately after a 30-min clinical protocol. The test uses intravenous boluses of 10 g glucose and 1 U ...insulin at t = 1 and 11 min, respectively, and measures glucose levels in samples taken at t = 0, 10, 20, and 30 min. The low clinical cost of the protocol is enabled via robust model formulation and a series of population-derived relationships that estimate insulin pharmacokinetics as a function of insulin sensitivity ( SI). Fifty individuals underwent the gold standard euglycaemic clamp (EIC) and DISTq within an eight-day period.SI values from the EIC and two DISTq variants (four-sample DISTq and two-sample DISTq30) were compared with correlation, Bland-Altman and receiver operator curve analyses. DISTq and DISTq30 correlated well with the EIC R = 0.76 and 0.75, and receiver operator curve c-index = 0.84 and 0.85, respectively. The median differences between EIC and DISTq/DISTq30 SI values were 13% and 22%, respectively. The DISTq estimation method predicted individual insulin responses without specific insulin assays with relative accuracy and thus high equivalence to EIC SI values was achieved. DISTq produced very inexpensive, relatively accurate immediate results, and can thus enable a number of applications that are impossible with established SI tests.
Background:
The pathogenesis of type 2 diabetes is characterized by insulin resistance and insulin secretory dysfunction. Few existing metabolic tests measure both characteristics, and no such tests ...are inexpensive enough to enable widespread use.
Methods:
A hierarchical approach uses 2 down-sampled tests in the dynamic insulin sensitivity and secretion test (DISST) family to first determine insulin sensitivity (SI) using 4 glucose measurements. Second the insulin secretion is determined for only participants with reduced SI using 3 C-peptide measurements from the original test. The hierarchical approach is assessed via its ability to classify 214 individual test responses of 71 females with an elevated risk of type 2 diabetes into 5 bins with equivalence to the fully sampled DISST.
Results:
Using an arbitrary SI cut-off, 102 test responses were reassayed for C-peptide and unique insulin secretion characteristics estimated. The hierarchical approach correctly classified 84.5% of the test responses and 94.4% of the responses of individuals with increased fasting glucose.
Conclusions:
The hierarchical approach is a low-cost methodology for measuring key characteristics of type 2 diabetes. Thus the approach could provide an economical approach to studying the pathogenesis of type 2 diabetes, or in early risk screening. As the higher cost test uses the same clinical protocol as the low-cost test, the cost of the additional information is limited to the assay cost of C-peptide, and no additional procedures or callbacks are required.
Despite the potential clinical benefits of normalizing blood glucose in critically ill patients, the risk of hypoglycemia is a major barrier to widespread clinical adoption of accurate glycemic ...control. To compare five glucose control protocols, a validated insulin‐glucose system model was employed to perform simulated clinical trials. STAR, SPRINT, UNC, Yale and Glucontrol protocols were assessed over a medical‐surgical intensive care unit patient cohort. Results were interpreted separately for patients with low to high sensitivity to insulin, and low to high variability in metabolic state. STAR and SPRINT provided good glucose control with risk of severe hypoglycemia less than 0.05% across all patient groups. UNC also achieved good control for patients with low and medium levels of insulin sensitivity (SI), but risk of severe hypoglycemia was raised for patients with high SI. Glucontrol showed degradation of performance for patients with high metabolic variability.
•The Gauss Newton parameter identification method was adapted to ignore outlier data.•The method is capable of determining outliers as a function of measurement variance.•The method was successfully ...tested in two models describing glycaemic dynamics.•Method application is not complicated and could thus be used in numerous scenarios.
The Gauss-Newton method is a simple iterative gradient descent method used to modify a mathematical model by minimising the least-squares residuals between the modelled response, and some observed behaviour. A common issue for parameter identification methods that optimise least-square residuals is the sporadic occurrence of outlying data in the observation data set.
This research proposes an amendment to the Gauss-Newton parameter identification approach that limits the influence of outlying data by dissipating the contribution of outlying data to the objective function that drives iterations. The modified method was tested in two and three-dimensional parameter identification exercises using virtual data from the dynamic insulin sensitivity and secretion test (DISST). The data incorporated random normally distributed noise (CV=3%) or random normally distributed noise in concert with an outlying data point. The proposed method performed similarly to the original method when no outlying data was included and found the model that fit accurately to the majority of data points when an outlying data point was present.
The proposed approach provides a valuable tool for the rejection of outlier data that is operator independent, does not require multiple stages of analysis, or manual removal of data.
► Insulin sensitivity broadly related to birth weight and gestational age. ► Relative change in insulin sensitivity not related to birth weight/gestational age. ► Prospective cohort similar in ...sensitivity to retrospective cohort. ► Control tightened using insulin sensitivities from insulin treatment periods.
Hyperglycaemia is a prevalent complication in the neonatal intensive care unit (NICU) and is associated with worsened outcomes. It occurs as a result of prematurity, under-developed endogenous glucose regulatory systems, and clinical stress. The stochastic targeting (STAR) framework provides patient-specific, model-based glycaemic control with a clinically proven level of confidence on the outcome of treatment interventions, thus directly managing the risk of hypo- and hyper-glycaemia. However, stochastic models that are over conservative can limit control performance. Retrospective clinical data from 61 episodes (25 retrospective to STAR, and 36 from a prospective-STAR blood glucose control study) of insulin therapy in very-low birth weight (VLBW) and extremely-low birth weight (ELBW) neonates are used to create a new stochastic model of model-based insulin sensitivity (SI L/mU/min). Sub-cohort models based on gestational age (GA) and birth weight (BW) are also created. Performance is assessed by the percentage of patients who have 90% of actual intra-patient variability in SI captured by the 90% confidence bands of the cohort based (inter-patient) stochastic variability model created. This assessment measures per-patient accuracy for any given cohort model.
Per-patient coverage trends were very similar between prospective and retrospective cohorts, providing a measure of external validation of cohort similarity. Per-patient coverage was improved through the use of BW and GA dependent stochastic models, which ensures that the stochastic models more accurately capture both inter- and intra-patient variability. Stochastic models based on insulin sensitivities during insulin treatment periods are tighter, and give better and safer glycaemic control. Overall it seems that inter-patient variation is more significant than intra-patient variation as a limiting factor in this stochastic forecasting model, and a small number of patients are essentially different in behaviour. More patient specific methods, particularly in the modelling of endogenous insulin and glucose production, will be required to further improve forecasting and glycaemic control.
Virtual trials have proved useful in developing safe and efficacious glycaemic control protocols. However, these trials rely on lumping all changes in patient condition into the insulin sensitivity ...parameter. As electronic data collection provides higher temporal resolution than paper-based charts, irregular timings of both therapies and measurements clash with a regular, hourly insulin sensitivity profile. Additionally, unobservable endogenous changes are a factor for hour-to-hour variability. This research extends the virtual trial protocol to natively handle irregular data by regularising the insulin sensitivity profile, and utilising a simple stochastic differential equation. The insulin sensitivity profile was re-interpreted as a b-spline basis, allowing a higher order description with greater local support. The fitting error resulting from this regularisation was absorbed by a stochastic element in the glucose compartment, representing the hour-to-hour changes that cannot be attributed to changes in insulin sensitivity. The resulting virtual patients were demonstrated to be equivalent to the originals when a 0th order basis was used. Inclusion of the stochastic element in this case simply ensured the model still fitted during periods of unmodelled high endogenous glucose production, while a 2nd order basis uses this element to natively control the balance between changes in patient state and hour-to-hour unmodelled changes due to noise and endogenous processes. The resulting virtual trials are thus better able to preserve information in irregular data sets, and regulate the balance between controllable and uncontrollable glycaemic changes.
Glargine and Glycemia: Pitfalls and Perils Fisk, Liam M.; Willis, Jonathan G.; Le Compte, Aaron J. ...
IFAC Proceedings Volumes,
2012, 2012-00-00, Letnik:
45, Številka:
18
Journal Article
Odprti dostop
Type 1 diabetics exhibit an unfulfilled basal insulin requirement, currently treated with long-acting subcutaneous insulins such as glargine. Due to glagine's unique flat peak the drug is an ideal ...basal insulin replacement. Use of the drug has extended beyond patients with diabetes, seeing use in critical care when patients are deemed stable but still require exogenous insulin. Data from four patients in the pilot trial of STAR in Christchurch hospital was gathered to outline serious considerations when using glargine in an ICU setting.
The patients were fitted with the ICING and Glargine Compartment models to identify time-varying insulin sensitivity (SI), which was plotted alongside the blood glucose (BG) trace, interstitial insulin compartment and insulin/nutrition inputs. Features of these profiles were then identified to elaborate on the risks associated with the use of a long-acting insulin analogue.
Importance of nutrition on patient safety, uncertainty in inter- and intra- patient variability in response to glargine doses, and time-scales of changes in patient condition were all highlighted from the four cases. The extended time-scale of physiological responses to glargine can put patients at risk of severe hypoglycemia if: A) metabolic condition changes dramatically within this period; or B) clinical limitations on nutrition are imposed after a dose is administered.
Although use of glargine has the potential to cater for patients with a basal insulin requirement and who have less requirement for intensive supervision, more research should be done into action of the drug in an ICU cohort before use becomes widespread.
Nutrition is an important factor in the treatment of patients in critical care. Potential hyper-rmetabolism means underfeeding may cause malnourishment, while overfeeding increases risk of ...hyperglycemia and the associated physiological impact. Hyperglycemia can be treated through accurate glycemic control (AGC), and this paper examines the link between nutrition and achievement of AGC. Clinically validated virtual trials were carried out on the 371 patients in the SPRINT cohort using STAR, an adaptive AGC protocol targeting 80-145mg/dL. Nutrition results were compared to the rates given clinically to investigate the effect modulating nutrition has on the final level of nutrition administered. The effect of clinical nutrition stoppages on this level of nutrition was also isolated. The link between nutrition and the ability to achieve AGC was investigated by targeting STAR to both 80-145mg/dL and 140-180mg/dL, allowing STAR to modulate nutrition as well as delivering constant rates of 60%, 80%, 100%, 120% and 140% ACCP goal. Performance was assessed as %BG within the target range, hyperglycemia as %BG above the range and clinical workload as the number of BG measurements. Relative tightness was estimated using BG IQR. As expected, modulating nutrition led to a range of total nutrition delivered to patients. Importantly, low nutrition administration corresponded to low insulin sensitivity, and clinical nutrition stoppages were shown to drop median nutrition rates by 10% over the first 4 days in ICU, suggesting a significant effect if a nutrition target is desired. Variable nutrition in STAR was shown to lead to reduced BG variability and clinical workload, and different nutrition rates showed significant differences in BG outcomes despite the adaptive STAR framework. Combined, these results show that AGC could be better achieved with less effort if variable nutrition was permitted. In part, this effect is due to constant nutrition restricting the ability of a protocol to respond to low insulin sensitivity. Constant nutrition will also have a strong effect on the ability to target a specific range.